Every good story has two sides, and in the case of Insurance Simulator 2026, those two sides are the why we built it and the how. Gavin has the why covered. This is the how. It was built in my own time, which means there were a lot of nights where I was staring at a screen at 3am, bleary eyed, wondering why on earth I was doing this to myself.
ResearchThe research phase, or “playing games and calling it work”
I grew up on simulation games. Football Manager, Theme Park, Theme Hospital, SimCity. More recently, The Roottrees are Dead, Return of the Obra Dinn and Papers, Please. That last group matters more than you’d think. They’re games about paperwork, deduction and spotting the thing that doesn’t add up. Which, if you squint, is insurance.
Each of those games lent something specific. Football Manager gave me the inbox-led UI, which seemed so very appropriate for insurance. An industry that runs on the inbox deserves a game that does too. Papers, Please shaped the claims process, where the fun comes from spotting the anomaly in a wall of otherwise plausible detail. Theme Hospital gave me permission to be silly. The Roottrees are Dead and Obra Dinn taught me that players will do a surprising amount of detective work if the payoff is a satisfying “aha”.
The toolset: Claude Code did the heavy lifting on the build itself, writing and rewriting the game with me across hundreds of sessions. Midjourney and Ideogram handled the art generation, Photoshop handled the character production line you’ll read about shortly, and my own imagination filled in the rest.
BuildBuilding in contact order
Like all good IT projects, I built it iteratively. I knew loosely where I wanted to end up, but I worked in what I started calling contact order: the order of the policy lifecycle itself. Underwriting leads to claims, claims lead to renewals. Build the front door before the back office. Yes, I know MTAs exist, but there’s only so much realism anybody wants!
The first job was the basics: found an insurer, receive submissions, quote, bind. Get that loop working end to end before touching anything else. It sounds obvious written down, but the temptation to wander off and build the fun stuff first is enormous, and resisting it is most of what “agile” actually means when you’re a team of one working at midnight.
The Andre “Clippy” tutorial arrived surprisingly early in that process. The game is complex enough that new players needed somebody holding their hand through the opening stretch, and it turns out a pixelated CEO popping up in the corner does that job rather well.
Throughout, I kept a running log of ideas. A lot of it came from the org chart. I knew which roles I wanted in the game, so the question became: how do you gamify each one? Unlock a CMO and you get a marketing budget with actual choices to make about where it goes. Unlock a CTO and you get to run an RFI procurement process, which I accept is a niche definition of entertainment, but this is a niche game. Marketing, technology, HR, claims. Each role needed a mechanism that genuinely affects play, not just a name on a chart.
The list still has plenty on it. Delegated authority, M&A, venture capital, global expansion, personal lines. Someone recently planted the idea that I started the game too late in the story, and that the real opening should see you as an underwriter at an insurer who eventually leaves to found their own. Maybe one day.
GamificationThe bit where it wasn’t fun
Here’s an uncomfortable truth about simulating insurance: if you do it accurately, you’ve built a job, not a game. The early build was very realistic. It was also not much fun, and far too much like guesswork. You’d receive a submission, stare at it, and have no real basis for a decision beyond vibes.
So there was a whole phase in the middle devoted purely to gamification. How do you make hiring interesting? How do you signal good and bad risks without just printing the answer on the screen? How do you make claims something other than admin? The answer, mostly, was layers. Layers of information that reward attention, and layers of interactivity that turn a decision into a moment. Claims became closer to Papers, Please, an anomaly-spotting exercise where something in the file doesn’t quite add up. Hiring picked up a tug-of-war mechanic, where rival insurers compete for the same candidate and you have to work out what actually matters to them from what you learned in the interview. There’s an XP and skill points system underneath it all, and yes, there are cheat codes. One player has already been caught and sacked by their board for fiddling the numbers, which I consider the game working exactly as intended.
That mid-development stretch was one of the most important of the lot. The little cinematic intro came late, just to give players some context before dropping them in.
UnderwritingThe underwriting problem
The bit I’m still struggling with, and I don’t pretend this is a finalised product ready for release on Steam, is the underwriting side. Because having been an underwriter, building this actually reminded me how little information underwriters are often working from. A list of sums insured, a target premium, a claims history and some risk features submitted on a form. There isn’t necessarily the inevitability and binary outcomes that a game perhaps requires.
It’s the main area I’d like to put more thought into: how to better gamify it, keep it fun, but encourage the basic principles of underwriting. It’s also the part where we need to go PROPERLY deep on the logic behind it. I’m talking split theft rates on non-ferrous sheet metal over certain limits due to theft practicality. That sort of thing.
Building it also reminded me how weird it was, in retrospect, how little feedback I got as an underwriter about performance. Honestly, I couldn’t tell you whether I spent the early years of my career as the single most profitable or most catastrophic underwriter in the industry. I wrote new business based on judgement and got very little (read: no) feedback on outcomes. It strikes me how useful that would have been to see tracked over time, gamified even, in competitive dashboards. Maybe that world has changed now.
Characters750 tiny insurance people
The most labour intensive part, by a distance, was the in-game characters. Version one used basic 16-bit sprites generated dynamically by the game itself. They were rubbish. Genuinely. So I moved to a layering system: base images with hair, beards, clothing and accessories stacked on top. Six skin tones, three age ranges spanning young, middle-aged and old.
You would not believe the detail that went into building a diverse, representative cast. Government data, population statistics, actual research, all to decide what a fictional claims handler looks like. Every layer then had to be created individually in Photoshop. Created, cut out, transparent background, naming convention. I set up batch jobs to recolour every hairstyle across the palette rather than repainting each one by hand. Some of the one-off styles even ended up with names. The whole time I’m picking up Photoshop skills and shortcuts I never knew existed, which is a strange side effect of building an insurance game.
Then came another micro-app. I loaded in the spreadsheet of characters and had it pick from the various layers to generate 750 unique character profiles, each named to match their demographics. Candidates even come with their own “LockedIn” profiles and CVs, because of course they do.
Shout out to Tony Canas, who played so extensively in the first week that I had to go back and patch in another thousand characters. He’d basically hired everyone. His all-conquering insurer broke my staffing model inside seven days.
I’m also aware of the sillier gaps. You can play for 200 years and nobody ages. If we ever fully productise the game, that sort of thing gets fixed. I’m equally conscious that not everybody may feel represented in the character set. Most of that comes down to technical limitations rather than intent, and we tried our best. There are a couple of unique characters in there already, and more representation may get patched in down the road. Create-your-character, interestingly, came much later, built from the same assets to let players make their own avatar.
PlaytestingPlaying it. A lot.
The biggest stage was playing it. Loads. Hundreds of hours of my own game before anyone else touched it.
The design principle that emerged from all that playing: you do something until it stops being fun, and just after that point, the game lets you hire someone to do it for you. Reviewing submissions is engaging for your first fifty. By your two hundredth, you want an underwriter. So the game gives you one, at roughly the moment tedium sets in. Delegation as a reward. Which is more or less how scaling a real business works. You don’t hire because a spreadsheet told you to. You hire because you cannot face doing that task one more time.
All that playtesting surfaced plenty that just wasn’t fun, which triggered a whole phase of quality of life fixes, bug hunting and features that existed purely to add depth or context. None of them individually matters much. Together they’re the difference between a demo and a game.
Since launch, other people have taken over the playtesting for me: circa 2,500 unique players and over 500 hours of combined playtime so far.
The main problem was knowing when enough was enough. There is no natural finish line on a project like this. Every session generates three new ideas, and the log of “things I’d like to add” grows faster than I can shrink it. I could happily plough another year of my life into this. It may end in divorce.
ReflectionsSo how hard was it, really?
Much easier than I expected. That’s the honest answer, and probably the most useful thing in this whole post. What it did need was a creative, problem-solving, logical brain to steer it. The vision has to come from you. But the tools aren’t just passive builders either. I constantly hit brick walls, and my move each time was to throw the problem straight back at the AI and ask for a few ideas on how to proceed. More often than not, one of them unstuck me. It became a genuine back and forth, and the tools are infinitely more powerful than you think.
They’re also comparatively cheap for what you’re getting. I built all of this within my existing Claude subscription. No extra tokens, no surprise bills, no infrastructure. A game with 2,500 players, built at the kitchen table on a standard subscription, by a marketer.
And the learning itself was half the reward. Photoshop batch processing, character pipelines, game design, none of which I knew a year ago. Picking up new skills at this stage of a career is fun and genuinely empowering, which is not something I expected to write about an insurance game.
So, got a problem? Try explaining it to Claude and see if it can help you build something to fix it. Worst case, you lose an evening. Best case, you’re the one writing one of these posts.